OpenAI released a big update today, ChatGPT-4 Turbo.
ChatGPT-4 Turbo provides more functionality, lower costs, and the ability coming soon to customize "GPTs" which will make it easier to deliver new AI capabilities to enterprises and governments.
You can now analyze an Excel file, upload a large PDF, create an image, and chat with an image from a single chat thread.
Resources
OpenAI Keynote (45 minutes)
Example How Apps Can Evolve with Assistants and Agents (minute 33:27 - 43:00)
Expanded Context Window (with better quality)
You can now "chat" with larger files (300 pages of a standard book).
The technical limit is 128,000 tokens compared to 8,000 tokens previously in ChatGPT-4.
More Control
JSON mode ensures you can integrate with applications more efficiently.
Predictable outputs can be made using "seed values."
Better World Knowledge
Knowledge of the world through April 2023
Browse the web, if needed.
New Modalities
Vision, text, and file analysis can be done in the same chat thread.
The API supports audio, text, and images - developers now have more tools to create new capabilities. This is a big deal.
Customization
Coming soon, you will be able to create custom GPTs for your business by uploading files.
Developers can build apps that work in ChatGPT similar to an app for an iPhone. Developers will make money in the GPT Store.
Higher Rate Limits
You can make more API calls which support AI applications that lean heavily on the large language model.
Copyright Shield
OpenAI will indemnify against copyright lawsuits.
Pricing
3x cheaper to prompt
2x cheaper for response (completion change)
About 2.75 x cheaper on average.
For Developers
Assistant API allows for building richer user experiences embedded in your apps.
Komentáře